Atomically-precise dopant-controlled single cluster catalysis for electrochemical nitrogen reduction

The ability to precisely engineer the doping of sub-nanometer bimetallic clusters offers exciting opportunities for tailoring their catalytic performance with atomic accuracy. However, the fabrication of singly dispersed bimetallic cluster catalysts with atomic-level control of dopants has been a long-standing challenge. Herein, we report a strategy for the controllable synthesis of a precisely doped single cluster catalyst consisting of partially ligand-enveloped Au4Pt2 clusters supported on defective graphene. This creates a bimetal single cluster catalyst (Au4Pt2/G) with exceptional activity for electrochemical nitrogen (N2) reduction. Our mechanistic study reveals that each N2 molecule is activated in the confined region between cluster and graphene. The heteroatom dopant plays an indispensable role in the activation of N2 via an enhanced back donation of electrons to the N2 LUMO. Moreover, besides the heteroatom Pt, the catalytic performance of single cluster catalyst can be further tuned by using Pd in place of Pt as the dopant.


Response:
We thank the referee for your valuable comments.
Typically, several parameters need to be taken into account for the evaluation of the overall catalytic performance for NRR. These include the NH 3 yield, faradic efficiency (FE) and the potential applied in electrochemical reaction. The optimized catalyst shall offer a high NH 3 yield with a high FE acquired at a low overpotential. In order to have a fair comparison, we need to consider all these factors to evaluate the overall catalytic performance of different catalysts.
We agree with referee that the FE value (~10% for Au 4 Pt 2 /G and ~12% for Au 4 Pd 2 /G) achieved in our experiment is not the highest one amongst all the reported values. However, these FE numbers are still considerably good compared to a majority of catalysts reported (refer to Figure  R1 and Table R1). Moreover, we also need to consider other two factors including maximal NH 3 production rate and the corresponding electrochemical potentials applied.
The ammonia production rate of Au 4 Pt 2 /G and Au 4 Pd 2 /G are 23.6 μg. ℎ .
Additionally, the best catalytic performance of these single cluster catalysts in terms of NH 3 yield and FE can be obtained at very low overpotential (-0.1 and -0.2 V versus RHE for Au 4 Pt 2 /G and Au 4 Pd 2 /G respectively). Therefore, the overall catalytic performance of both Au 4 Pt 2 /G and Au 4 Pd 2 /G catalysts reported in this work is superior. Moreover, this work demonstrates that the heteroatom dopant plays an indispensable role in the activation of N 2 . Besides the heteroatom Pt, the catalytic performance of bimetallic SCCs can be further tuned by using Pd in place of Pt as the dopant. Our findings offer a new route for the design of novel SCCs with atomic precision for electrochemical dinitrogen reduction. Figure R1 (Supplementary Fig. 1). NRR catalytic performance of the Au 4 Pd 2 /G and Au 4 Pt 2 /G catalysts in comparison with the catalysts reported. Response: We thank the reviewer for your suggestion. Two bimetallic clusters synthesized in this work (Au 4 Pt 2 and Au 4 Pd 2 ) share the same octahedral framework. This is crucial to establish the composition-property correlation by a direct comparison of their electrochemical NRR performance. Unfortunately, pure Au or Pt clusters (e.g. Au 6 or Pt 6 ) with the same octahedral framework as that of Au 4 Pt 2 have not been obtained up to date. In light of referee's comments, we have tested the NRR performance of Au 25 and a newly prepared Pt cluster (with an average size of 1 nm). The results ( Figure R4) reveal a poorer NRR performance of both pure Au and Pt clusters as evidenced by a lower yield and lower faradic efficiency compared to that of Au 4 Pt 2 and Au 4 Pd 2 . Therefore, these results further confirm that hetero-dopant (Pt or Pd) of bimetallic clusters play importance roles in the enhanced catalytic performance of NRR. Fig. 18) TEM images of pure Au and Pt cluster catalysts (note that both Au and Pt clusters are also anchored on the defective graphene) Figure R5 (Supplementary Fig. 19). Catalytic performance of pure Au and Pt clusters catalysts in ENRR.

Figure R4 (Supplementary
Action: We have placed Figure R4 and Figure R5 with the corresponding description in the revised Supplementary information.

Reviewer #3 (Remarks to the Author):
In this study, the authors presented a strategy for the synthesis of cluster catalysts consisting of partially ligand-enveloped Au4Pt2 and Au4Pd2 clusters on graphene. And the suggested nanoscale confined interfacial space between the graphene substrate and bimetallic cluster is regarded as the active site for N2 fixation. Although authors design atomically precise bimetallic cluster through simple experimental operations, some indispensable experiments and reasonable explanations are missing at this stage. To be fair, the electrochemical synthesis of NH3 is a burgeoning research field, thus a rigorous refereeing is needed. Based on the overall evaluation of the article, the manuscript did not meet the standard of Nat. Comm., thus the manuscript should be rejected at this stage.
Comment 1. The authors should examine the reducing regent for the synthesis of ultrafine Au-Pt bimetal clusters by using 2-phenylethanethiol (Page 5) or thiol (Page 4).

Response:
We apologize for the confusion and would like to clarify this point as follows.
2-Phenylethanethiol is considered as one specific type of thiol. In page 4, we use "thiol" because it refers to a general concept. In page 5, we specify the thiol used in our work is 2phenylethanethiol. We would like to point out that only 2-phenylethanethiol is used as the reducing regent in our work. Comment 2. The authors described "…an intense peak at m/z ~2274 is observed, which can be assigned to the molecular ion of the bimetal cluster…" (Page 5). A detailed analysis should be performed.

Response:
We appreciate the reviewer's suggestion. In view of referee's comments, we have provided a detailed analysis of MS spectrum in our revised supporting information.
Mass spectroscopy (MS) and TGA are two commonly-used characterization techniques for metal clusters. High resolution MS provides the accurate analysis of molecular mass of molecular-like clusters. TGA will offer additional information related to the mass contribution of ligand involved in the cluster.
The molecular weight of as-prepared Au 4 Pt 2 (SC 2 H 4 Ph) 8 cluster is determined to be 2274 Da (the peak with the largest M/Z value as shown in Figure 1a).
The mass contribution of ligand is determined to be 48% based on the TGA result. Therefore, the mass of ligand can be calculated as 2274 × 48% = 1091.5 Da. The total number of ligand can be  Au 4 Pd 2 /G catalyst yields a NH 3 production rate of 13.1 µg mg -1 h -1 at -0.1V, lower than that of Au 4 Pt 2 /G catalyst at the same potential. This indicates that Au 4 Pd 2 /G has a lower ENRR activity compared to Au 4 Pt 2 /G. However, we obtained a maximum NH 3 yield rate of 27.1 µg mg -1 h -1 with a FE of ~12% at a more negative potential of -0.2 V for Au 4 Pd 2 /G, actually outperforming the Au 4 Pt 2 /G (Fig. 3) at a more negative potential.
These observations suggest that hydrogen evolution reactions (HER) could be more effectively suppressed in this system as compared to that of Au 4 Pt 2 /G. This is also consistent with the fact that Pt generally favors the HER, which severely limits ENRR towards NH 3 production at more negative potentials. To verify this, we have evaluated the HER performance of both Au 4 Pd 2 /G and Au 4 Pt 2 /G catalysts (Supplementary Fig. 17). The results clearly demonstrate that the overpotential of Au 4 Pd 2 /G for HER is higher (more negative) than that of Au 4 Pt 2 /G, suggesting the competing HER reaction has been more effectively suppressed in the process of ENRR.
[1] We have provided the explanation for a higher NH 3 yield and faradic efficiency for Au 4 Pd 2 /G catalyst at more negative potential in the main text.
in page 12: "we obtained a maximum NH 3 yield rate of 27.1 µg mg -1 h -1 with a FE of ~12% at a more negative potential of -0.2 V for Au 4 Pd 2 /G, actually outperforming the Au 4 Pt 2 /G (Fig. 3). This suggests that HER could be more effectively suppressed in this system as compared to that of Au 4 Pt 2 /G, consistent with the HER performance of two bimetallic SCCs tested" [2] A short description of HER performance of both catalysts have been included in the revised supporting information. Comment 4. The authors addressed "the 1D polymeric cluster chain was observed to disassemble into individual clusters upon dissolving in organic solvents" (Page 6). However, the experimental evidence is missing.
Response: Thanks for the reviewer's comment. We disagree with the reviewer regarding this point. The following experimental evidence proves the disassembly of 1D polymeric chain in the organic solvents.
[1] Firstly, the mass spectrum reveals a molecular ion peak (the signal of the largest M/Z value) at 2274 Da (Fig. 1a). This proves the dissociation of polymeric chain into individual clusters, which otherwise cannot produce such a unique feature peaked at 2274 Da [2] One dimensional chain-like structure is expected to form on graphene if 1D polymeric chain remains intact. This is in contrast to what we have observed in TEM and AFM images (  Response: Thank the reviewer for pointing this out. The measurement temperature was not intentionally changed for two different crystals. In light of your comments, we have carried out SXRD measurement of Au 4 Pt 2 (SR) 8 at 100 K. The corresponding two sets of data for both crystals measured at the same temperature are included in the revised Supplementary information.

Comment 7. On Page 9 (Supplementary Information), the thermogravimetric analysis result of Au4Pt2(SR)8 crystals. The curve showed two thermal decomposition processes, and the related reaction process should be answered.
Response: This is a good point. We would like to provide the following discussion for the explanation of our TGA result.
The first weight loss of (~37%) can be attributed to the removal of eight -C 2 H 4 Ph organic part from thoil ligand (-S-C 2 H 4 Ph), and the subsequent weight loss (~11%) stems from the desorption of eight sulphur atoms of metal cluster. A complete loss of eight -SC 2 H 4 Ph during TGA measurement results in a total weight loss of 48% in the whole process ( Supplementary Fig. 2). Fig. 2). Two-step weight loss with the corresponding percentage.

Comment 8. In figure 2(f), it is difficult to distinguish the arrangements of Au or Pt atoms (the plotting scale notes are missing), some high-resolution pictures should be provided.
Response: We thank the referee for this comment.
All the scale bars in Figure 2f are 5 Å. It has been included in the figure caption of 2f in the revised manuscript. Figure 2f presents the high-resolution STEM images of anchored Au 4 Pt 2 clusters on graphene.
The results reveal that majority of bright dots contain a cluster of six atoms, as expected for the bimetallic Au 4 Pt 2 cluster. The atomic arrangements of the imaged clusters vary, which can be attributed to the different viewing direction of the clusters or electron-beam induced cluster dissociation. Due to the close atomic mass between Pt and Au, it remains a great challenge to distinguish these two atoms based on the image contrast of STEM.

(Supplementary Information), what is the concentration ratio of 14NH4+ and 15NH4+ in NMR, whether the value is proportional to the input amount of 14N2 and 15N2 molecules.
Response: This is a good point.
The integrated peak area associated with 15 NH 4 + and 14 NH 4 + is determined to be 1:9, which is proportional to the input ratio between 15 N 2 and 14 N 2 molecules.

Comment 10. The ammonia yields and Faradic efficiencies of multiple cycling tests should be tested.
Response: This is a good suggestion. In view of referee's suggestion, we carried out the multiple cycling tests of both Au 4 Pt 2 /G and Au 4 Pd 2 /G catalysts for ENRR.
As shown in Figure R2, the results reveal that both ammonia yield and FE remain nearly constant during the multiple cycling stability tests. Both catalysts demonstrate excellent stability during the process of electrochemical N 2 reduction reaction, see Figure R2 for details. Figure R2 (Supplementary Fig. 20). Long-term stability of both Au 4 Pd 2 /G and Au 4 Pt 2 /G catalysts for NRR.
Action: We have included the stability test results in the revised Supplementary information.

Comment 11. The structural characterization of bimetallic cluster after electrochemical N2 reduction reaction should be provided.
Response: We thank reviewer for this valuable comment. We have conducted additional experiment to characterize the structure of the Au 4 Pt 2 /G and Au 4 Pd 2 /G after electrochemical reaction.

Figure R6
presents TEM images of both Au 4 Pt 2 /G and Au 4 Pd 2 /G catalysts before and after NRR reactions. It is observed that the cluster on graphene shows negligible aggregation and thus the size of these clusters remains nearly constant before and after reaction.
Spherical aberration corrected STEM images of both catalysts show that the clusters still consist of six-atom after NRR reaction, thus the cycling stability of both catalysts was further confirmed, (see Figure R6).
Additionally, XAFS measurement on Au 4 Pt 2 /G and Au 4 Pd 2 /G catalysts after ENRR was also performed. As shown in Figure R7, the Au, Pt L 3 -edges and Pd K-edge show negligible changes after ENRR. This provides a clear evidence to support that the catalyst remains stable during the reaction. Figure R6 (Supplementary Fig. 21). Structural characterization of bimetallic cluster after electrochemical N 2 reduction. a, b TEM images of Au 4 Pt 2 /G before and after ENRR. c, d TEM images of Au 4 Pd 2 /G before and after ENRR. (Inset b and d are spherical aberration corrected STEM images of Au 4 Pt 2 /G and Au 4 Pd 2 /G catalysts after ENRR. Scale bars for inset are 5Å) Figure R7 (Supplementary Fig. 22) XANES spectra of Au 4 Pd 2 /G and Au 4 Pt 2 /G catalysts before and after ENRR. (a) Pd K-and (b) Au L 3 -edges XANES spectra of Au 4 Pd 2 /G catalyst, (c) Pd K-edge and (d) Au L 3 -edge XANES spectra of Au 4 Pd 2 /G catalyst before and after ENRR, (e) Pt K-edge and (f) Au L 3 -edge XANES spectra of Au 4 Pt 2 /G catalyst before and after ENRR.

Response:
We appreciate the reviewer for the constructive comment.
In view of referee's comments, we have carried out modelling of additional structures (Au 4 Pt 2 (SR) 4 /G, Au 4 Pt 2 (SR) 2 /G and Au 4 Pt 2 /G) suggested by referee. The corresponding simulated XANES spectra of these structures are shown in Figure R9 below. Amongst all the structures proposed, the simulated spectrum of Au 4 Pt 2 (SR) 6 /G still show the best match with experimental data. Hence, our conclusion on the proposed structure (Au 4 Pt 2 (SR) 6 /G) is still valid. Figure R9 (Supplementary Fig. 10). A comparison of experimental XANES spectra with the simulated spectra of Au 4 Pt 2 (SR) 4 /G, Au 4 Pt 2 (SR) 2 /G and Au 4 Pt 2 /G. Response: Generally, it is very difficult to model solvation effect on the reaction barriers by ab initio calculations due to the complexity of solve involved systems. In light of your comment, we used a simple method to estimate the order of the solvation effects on the energies of the ratelimiting step, i.e. the NH 3 desorption (as shown in Fig. R10 below in the response to Comment 14 or Fig. 5d in the revised manuscript): We calculated the energy barriers of this ratedetermining step with/without one water molecule in the vicinity of NH 3 molecule. Our calculations show that the barrier increases slightly by 0.06 eV with the water molecule. This result suggests that the solvation effect in the system is not significant and the major conclusion of the paper is still valid.

Comment 14.
In this study, the N2 adsorption step is favored over the catalysts, while the desorption process of *N is not provided, and subsequent steps should be added in Figure 5d.

Response:
We appreciate the referee's constructive comments. We agree with referee that a full reaction pathway calculation is needed to provide a comprehensive understanding of the reaction mechanism. We thus carried out additional calculation of the whole reaction pathway of N 2 transformation. Our results (Figure R10 (Fig. 5d)) show the full reaction path for the pathway I (in which the protonation of N bonded graphene occurs first) is a more favourable route. Note that with more accurate calculations, we are able to locate a new stable state for the 4 th step (2 nd H in the figure) which is energetically more stable than the one showed in the previous submission. We can see from the figure that the rate-limiting step is indeed the desorption of NH 3 (the final step) with a maximum barrier of ~0.91 eV. In Figure R11 (Supplementary Fig.  14), we present the full reaction path of the other mechanism in which the N atom binding with the metal cluster leaves first (the path II in the previous submission). For this mechanism, there are two rate-limiting steps: (i) the desorption of the first NH 3 with a barrier of 0.74 eV; (ii) the formation of the second NH 3 with a high barrier of 2.42 eV. Therefore, it is most likely the reaction proceeds along the pathway I based on these calculation results. Figure R10 (Fig. 5d). Calculated energy profile of the reaction pathway for the pathway I in which the protonation of N atom bonded graphene occurs first. Note that NH 3 form when the 3 rd and 6 th H added. Figure R11 (Supplementary Fig. 14) Calculated energy profile of the reaction pathway for the alternative mechanism (pathway II in the previous submission). For this mechanism, there are two rate-limiting steps: (i) the desorption of the first NH 3 with a barrier of 0.74 eV; (ii) the formation of the second NH 3 (after the 6 th H added) with a high barrier of 2.42 eV.
Comment 15. The PDOS of different orbitals for N2 adsorbed over the catalyst has been discussed in the manuscript. As an important step, the PDOS results of catalysts with *NNH should also be added in the revised manuscript.
Response: In light of referee's comment, we calculated the PDOS of *NNH generated in the pathway I. Figure R12 reveals a significant hybridizations between N and both C and H that are bonded with N 2 , indicating that N 2 forms chemical bonds with both C and H. The formation of these bonds is energetically favourable according to the energy profile shown in Fig. R10, which is due to the fact that the N 2 molecule is already activated when adsorbed on the catalyst. Figure R12 (Supplementary Fig. 15). Calculated PDOS of *NNH produced for the pathway under distal mechanism: (a) PDOS for all chemical species and (b) Enlarged PDOS for N, C and H that are bonded with N 2 . For a better visualization, the PDOS of H has been enhanced by 10 times.
Action: Figure R12 with the corresponding description has been included in the revised supporting information.
The re-submitted work by Yao et al. reported electrocatalytic nitrogen reduction under ambient conditions over obtained Au4Pt2 and Au4Pd2 clusters on graphene. Although the revised draft showed an improvement over the previous one, some indispensable experiments should be added, because the authenticity of electrocatalytic ammonia synthesis has been questioned by many scholars (Nature, DOI: 10.1038/s41586-019-1260-x). Therefore, this manuscript is not recommended for publication in the present form. If the following questions can be answered satisfactorily, the manuscript will be reconsidered. 1. Although the authors conducted a mixture of 14N2 and 15N2 labeling to confirm the source of nitrogen for NH3 production in Supplementary Fig. 7 Supplementary Fig.  19(b). The reviewer expect the authors to use surface-enhanced infrared absorption spectroscopy to study the reaction mechanisms of NRR over obtained Au4Pt2 and Pt clusters. 3. The maximum nitrogen production rate is up to 23.6 µg mg-1 h-1 over Au4Pt2/G in this manuscript. The corresponding current density is not reported, nor the rate normalized to either geometric or electrochemical surface area. The actual catalyst loading mass does not have provided, so these important metrics cannot even be computed from the current data provided. 4. It is important to evaluate how much N2 was reduced to NH3 during the electroreduction process (so the amount of N2 conversion rate can be evaluated). Please describe the testing process in detail. 5. Some spelling and expression errors need to be corrected, such as, "In contrast, the Au4Pt2/G SCC generates a maximum NH3 yield of up to 23.6 µg mg-1 h-1 at 0.1 V (should be -0.1V)", "Our findings have carved out…".
Reviewer #4 (Remarks to the Author): Single-cluster catalysis (SCC) is an emerging and exciting topic in heterogeneous catalysis, but it remains a grand challenge to synthesize SCC with atomic precise. In this respect, this work represents a breakthrough. Moreover, the Au4Pt2 and Au4Pd2 SCC show promising performances in the very demanding reaction of electrochemical nitrogen reduction. I believe this work is of highly novelty and importance, and therefore can be acceptable for publication in Nature Communication after addressing the following concerns: 1. EXAFS fitting results need to be given to see if the Au-S and Pt-S distances as well as coordination numbers are consistent with the other characterization and theoretical results. 2. A blank test with defective graphene, instead of non-defective graphene, needs to be performed to see if this support itself is active for NRR. 3. The authors claim that the interfacial space between the graphene substrate and the Au4Pt2 cluster acts as the active site. What does the interfacial space? And what is difference between the interfacial space and interface? 4. From the XANES spectra, both Pt and Au atoms in the cluster catalyst are positively charged, in this case, it might not be easier to transfer electrons to N2 molecules in comparison with negatively or zero charged metallic clusters. Please comment on this point.

Reviewer #3 (Remarks to the Author):
The re-submitted work by Yao et al. reported electrocatalytic nitrogen reduction under ambient conditions over obtained Au4Pt2 and Au4Pd2 clusters on graphene. Although the revised draft showed an improvement over the previous one, some indispensable experiments should be added, because the authenticity of electrocatalytic ammonia synthesis has been questioned by many scholars (Nature, DOI: 10.1038/s41586-019-1260. Therefore, this manuscript is not recommended for publication in the present form. If the following questions can be answered satisfactorily, the manuscript will be reconsidered.

Response:
We thank the referee for your positive comments on the revised manuscript. Supplementary Fig. 7, it's not precise enough. Aiming toward higher accuracy and reproducibility of ENNR results, a thorough discussion of experimental parameters and testing methods in ENRR should be added, such as the reported protocols in Nature, DOI: 10.1038/s41586-019-1260-x. Chem. Soc. Rev., DOI: 10.1039/c9cs00280d. ACS Catal., 2018, 8, 7820-7827. Nature Catalysis, 2019. Additionally, a rigorous test procedure must be put forward that, by making essential use of 15N2, allows the public to reliably detect and quantify the electroreduction of N2 to NH3.

Response:
We thank the referee for valuable comments. We noted that several groups have published their recent work on how to design and evaluate ENRR with protocols for the precise determination of ammonia yield. These protocols allow to verify the source of ammonia. We certainly agree with the referee it can exclude the possible background/extraneous ammonia sources for the precise determination of ammonia production if we apply the same protocol published. Unfortunately, the extremely high cost of 15 N2 gas is not affordable for us to adopt the exact protocol published. In view of referee's comments, we developed a cost-effective alternative approach to further rule out the possible background/contamination ammonia sources in our experiment. The additional measurements/results prove that the data is valid and the ammonia is produced by ENRR process. We will describe our new approach and present additional data as follows.
The key of our method lies in the variation of the ratio of starting 14 N2/ 15 N2 gas and check the corresponding 14 NH4 + / 15 NH4 + generated after ENNR. If the correlation between these two cases can be established, this would provide a compelling evidence to support the validity of the yield and FE of ammonia generated by ENRR.
In our initial manuscript, we have conducted the 15 N2 labeling experiment to confirm the source of nitrogen for NH3 production. Instead of the use of pure 15 N2 gas (due to the high cost), we mixed 15 N2 and 14 N2 with a mole ratio of 1:9 ( 15 N2/ 4 N2) to confirm the NH3 production. The ratio of 15 NH4 + and 14 NH4 + is also determined to be 1:9 (obtained from NMR signals associated with the 15 NH4 + and 14 NH4 + ), proportional to the input gas ratio of 15 N2/ 4 N2 (see Figure R1 and Supplementary Fig. 7 of our original Supplementary Information). This confirms that NH3 is indeed generated by ENNR process.
To further validate this approach, we preformed additional experiment by varying the ratio of 15 N2/ 14 N2. In this round of experiment, a ratio of 1:1 ( 15 N2/ 14 N2) was utilized as the feed gas for electrochemical reduction. The ratio of 15 NH4 + / 14 NH4 + produced was determined approximately to be 1:1 using 1 H-NMR ( Figure R2). All these results point out that the ammonia obtained in our experiment is derived from dinitrogen reduction rather than from the nitrogen impurities or environmental contaminations. The experimental parameters and testing methods for ENRR as well as the details of 15 N2 labeling test were included in our revised manuscript (see Method section 7-13 of the Supplementary Information).

Actions:
 In light of referee's comments, we included the detailed information regarding experimental parameters and testing methods for ENRR in the revised Supporting Information (highlighted with yellow background).  We also include the additional data on 15 N2 labelling ( Figure R2) in the manuscript and merge it into Fig 3 of the main text as panel d. Figure R1: Supplementary Fig. 7 1 H-NMR spectra of 14 NH4 + and 15 NH4 + produced from ENRR using (a) 14 N2, and (b) a mixture of 14 N2 and 15 N2 ( 15 N2 enrichment ~10%). The integrated peak area associated with 15 NH4 + and 14 NH4 + is determined to be 1:9, which is proportional to the input ratio between 15 N2 and 14 N2 gas. Figure R2: 1H-NMR spectrum of 14 NH4 + and 15 NH4 + produced from ENRR using a mixture of 14 N2 and 15 N2 ( 15 N2 enrichment ~50%). The integrated peak area associated with 14 NH4 + and 15 NH4 + is determined to be 1:1, which is proportional to the input ratio between 14 N2 and 15 N2 gas. and 15 N2 gas.

It has been reported that no absorption band associated with N was observed on Pt surfaces in liquid via surface-enhanced infrared absorption spectroscopy (J. Am.
Chem. Soc., 2018Soc., , 140, 1496Soc., -1501. But in this manuscript, the authors reported Pt cluster showed the ability of NRR in Supplementary Fig. 19(b). The reviewer expect the authors to use surface-enhanced infrared absorption spectroscopy to study the reaction mechanisms of NRR over obtained Au4Pt2 and Pt clusters.

Response:
We appreciate the reviewer for this good suggestion. This is a highly relevant reference (J. Am. Chem. Soc., 2018, 140, 1496-1501. The authors demonstrate a nice mechanistic study of ENRR using surface-enhanced infrared absorption spectroscopy. However, Pt in the Au4Pt2 cluster (ionic nature, positive charge) is distinct from metallic Pt foil used in the reference. Therefore, they show different electrocatalytic activities.
In line of reviewer's suggestions, we also attempted to use the surface-enhanced infrared absorption spectroscopy (SEIRAS) to study the mechanism of the reaction. The results obtained is shown in (Figure R3). The absorption signals related to H−N−H bending, −NH2 wagging, of adsorbed N2Hy species are expected to occur around 1450, 1298, 1109 cm -1 respectively. In this work (J. Am. Chem. Soc., 2018, 140, 1496-1501, the authors studied the N absorption on a large-area Au (or Pt) thin foil (1.76 cm 2 ) with a much higher mass of Au (or Pt) compared to that of Au (or Pt) used in our system. This leads to a much prominent signal of N-contained species. However, the mass of Au (or Pt) of the Au4Pt2/G (Pt/G) catalyst used for ENRR is in the microgram scale (noted 1 mg of Au4Pt2/G with the Au4Pt2(C2H4Ph)8 loading of 8.5 wt% was used in the experiment). If we exclude the mass from the ligand (-C2H4Ph) of the cluster, the actual loading of Au and Pt is lower than 3 wt% and 1.5 wt%, respectively). Because of this, in our experiment, the signal between ∼1300 and 1500 cm -1 is rather featureless. It thus remains very difficult to compare the distinct IR features during ENRR. Nevertheless, we would like to mention that major focus of our work lies in the design and synthesis of atomically-precise ultrafine bimetal SCC for ENRR. A particularly interesting aspect of our findings is that the heteroatom dopant is revealed to play an indispensable role in the ENRR. The ability to precisely dope SCC raises the prospect of designing a wide range of atomically precise SCCs with dopant-controlled reactivity for broad applications beyond NH3 production.

Actions:
We have cited the reference suggested by referee in the revised introduction part of our manuscript 3. The maximum nitrogen production rate is up to 23.6 µg mg-1 h-1 over Au4Pt2/G in this manuscript. The corresponding current density is not reported, nor the rate normalized to either geometric or electrochemical surface area. The actual catalyst loading mass does not have provided, so these important metrics cannot even be computed from the current data provided.

Response:
We thank the reviewer for these valuable comments. Sorry for the confusion. Some of these details were provided in the original supporting information. In view of the reviewer's comments, we have provided the yield rate (normalized by geometric area) of ammonia over Au4Pt2/G catalyst and loading of the metal clusters on graphene in the revised supporting information.
[1] The current (by integrating Chronoamperometric curves of an electrochemical reaction) and the area of the carbon paper (22 cm -2 ) were included in our supporting information ( Supplementary Fig. 6a, 6c and Part 8 of Method). Therefore, the current density can be readily calculated based on these details provided. We presented current rather than current density in the report because for the calculation of Faradaic efficiency (FE) of an electrochemical reaction, the electrode area is not needed according to the formula below where F is the Faraday constant (96485 C mol -1 ) and 3 represents the mole of ammonia; Q is the total charge passed through the electrode [Current (A)×Time (S)]. [2] In some of the previous reports, the authors reported the value normalized by the geometric area, particularly for some non-noble metal catalysts. However, we noted that the mass normalization is also widely used in some reports, in particular, for the evaluation of the electrocatalytic performance of noble metal based nanocatalysts (e.g. Nat. Commun.| , 9:1795. In our work, the mass normalized yield rate was adopted to compare the catalytic performances between Au4Pd2/G and Au4Pt2/G catalysts, see Table R1. [3] In line of referee's suggestions, we also determined the yield rate of ammonia normalized by geometric area (e.g. Au4Pt2/G as catalyst). The corresponding results were also shown in Table R2 and Figure R4. We also provided the loading of metal clusters on graphene is 8.5 wt% for Au4Pt2/G and 10.5 wt% for Au4Pd2/G catalyst in the revised supporting information.

Action:
 We have included Table R1 and Table R2 in the revised Supplementary information (see Supplementary Table 3 and Supplementary Table 4 for details)  The loadings of metal clusters were supplemented in the revised Supporting information (part 8, the preparation of cathode for ENRR).
8. The preparation of cathode for ENRR. Typically, 1 mg catalyst (the loading of metal clusters on graphene is determined to be 8.5 wt% for Au4Pt2/G and 10.5 wt% for Au4Pd2/G, respectively) and 5 μL of Nafion solution (5 wt%) were dispersed in the absolute ethyl alcohol (100 μL) followed by the sonication for 30 minutes to form a homogeneous ink.  Figure R4. The ammonia yield rate normalized by the geometric area (Au4Pt2/G).

It is important to evaluate how much N2 was reduced to NH3 during the electroreduction process (so the amount of N2 conversion rate can be evaluated).
Please describe the testing process in detail.

Response:
We thank reviewer for this comment. Based on the experiment we conducted, the N2 conversion rate is quite low (～3.5×10 -5 %). In the majority of previous reports on ENRR (including our experiment), N2 gas is supplied under a continuous ventilation because of the lack of a recycle setup in the experiment so that the precise N2 conversion cannot be determined with high accuracy.
Due to an extremely low yield of NH3, it also remains challenging to precisely determine the N2 conversion rate as the accuracy of reported value varies from case by case. If one wants to obtain a higher N2 conversion rate, it is required to reduce N2 flow and ensure the saturation of N2 in the electrolyte (note nitrogen can be easily saturated in aqueous electrolyte).
We would like to emphasize that the major focus of the current research (including our work) lies on the design of new efficient catalysts and fundamental study of the mechanistic insights. There is still a long way ahead towards the real-life application of electrochemical N2 reduction.

Actions:
In view of referee's comments, we have provided the testing process in details in the revised supporting information (7-13 of Method part) 5. Some spelling and expression errors need to be corrected, such as, "In contrast, the Au4Pt2/G SCC generates a maximum NH3 yield of up to 23.6 µg mg-1 h-1 at 0.1 V (should be -0.1V)", "Our findings have carved out…".

Response:
We thank the reviewer for pointing out. We have revised these errors/typo in the revised manuscript.
…This work provides a new route for the design of novel SCCs with atomic precision for broad applications beyond NH3 production.
In contrast, the Au4Pt2/G SCC generates a maximum NH3 yield of up to 23.6 µg mg -1 h -1 at -0. Au-S 2 2.33 species in the cluster will behave very differently from the negatively charged or zerovalence metal clusters. If the charge transfer is mainly governed by electrostatic interactions between metal ions and electrons, we agree with referee that it would be harder to transfer electron from positively charged cluster to N2 as compared to negatively or zero charged metallic clusters. However, in the case studied here, the clusters are supported on graphene. The charge transfer between metal atoms and the N2 molecule is mainly controlled by the hybridization between metal and N2 orbitals, for which the orbital alignment (rather than the electrostatic interaction) is a key factor. Our calculations (Fig. 5b and Fig. S13) have revealed that the Fermi level of graphenesupported metal clusters is very close to the LUMO of N2 molecule so that the hybridization between N2 LUMO and occupied metal states become possible, leading to the charge transfer. Figure R1. FTIR spectra of Au4Pt2(SR)8. The signals between ∼1250 and 1500 cm -1 , overlap with the absorption signals related to adsorbed N2Hy species. Figure R2. FTIR spectra during the first segment from 0.15 to -0.6 V (vs. RHE) on the Au4Pt2(SR)6/G electrode in a N2-saturaed 0.1 M HCl solution.
In view of referee's comments, we also conducted the electrochemical N2 reduction using a home-design equipment under a closed nitrogen-saturated environment (The N2 volume is 2 L). The concentration of NH4 + was determined as 5.31 μg/mL by using indophenol blue method ( Figure R3) after the reaction has been carried out for four days. The total volume of electrolyte is 30 mL and thus the total mass of NH4 + is calculated to be 159.32 μg. In this case, the N2 conversion rate is determined to be ~ 4.96×10 -3 % (Table R1). Figure R3. UV-Vis absorption spectrum of the electrolyte after performing ENRR for four days using Au4Pt2/G catalyst (indophenol method. Inset: cuvettes with electrolyte with/without electrochemical reaction).
2. Due to the severe deviation of the AFM test baseline, the author should retest the particle size in Supplementary Fig. 5.

Response:
We have conducted AFM measurement to better estimate the cluster size (height) according to referee's suggestion.
AFM image in (Figure R4, also Supplementary Fig. 5) reveals the morphology of